Differences with Apache Hive

• Completely programmatic, which means it's got a lot more dynamic/meta-capabilities.

• No temporary tables; all tables are basically 'external'.

• No persistent metadata repository; the schemas (DDL) all in code.

• It's not possible to sort group-by columns in descending order (this appears to be a Cascading/Hadoop limitation due to fields being sorted in natural order prior to the reduce phase).